From: Ross Johnson <rpj@ise.canberra.edu.au>
To: Michael Johnson <michaelj@maine.rr.com>
Cc: pthreads-win32 discussion list <pthreads-win32@sources.redhat.com>
Subject: Re: Deadlock interaction between pthread_cond_check_need_init.c and pthread_cond_destroy.c
Date: Thu, 19 Sep 2002 19:40:00 -0000 [thread overview]
Message-ID: <3D8A8A88.3000109@ise.canberra.edu.au> (raw)
In-Reply-To: <3D89F90A.2030805@maine.rr.com>
Michael,
I've applied your patch and committed it to CVS. I'll write a test
for it later.
Many thanks for this.
Ross
Michael Johnson wrote:
> When two different threads exist, and one is attempting to destroy a
> condition variable while the other is attempting to initialize a
> condition variable that was created with PTHREAD_COND_INITIALIZER, a
> deadlock can occur:
>
> Thread A Thread B
> Enters ptw32_cond_check_need_init
>
>
> Enters pthread_cond_destroy
> EnterCriticalSection(&ptw32_cond_test_init_lock)
>
> (now holds ptw32_cond_test_init_lock)
> EnterCriticalSection(&ptw32_cond_list_lock)
> Enters pthread_cond_init
> Determines that condvar is static initialized
> EnterCriticalSection(&ptw32_cond_list_lock)
> EnterCriticalSection(&ptw32_cond_test_init_lock)
> (now waiting to enter ptw32_cond_list_lock)
> (now waiting to enter ptw32_cond_test_init_lock)
> deadlocked
>
>
> It appears from reading the code and from a patch that I made that the
> following change to pthread_cond_destroy appears to fix this:
> |
> |
>
> |{|
> | pthread_cond_t cv;|
> | int result = 0, result1 = 0, result2 = 0;|
> ||
> | /*|
> | * Assuming any race condition here is harmless.|
> | */|
> | if (cond == NULL|
> | || *cond == NULL)|
> | {|
> | return EINVAL;|
> | }|
> ||
> | if (*cond != PTHREAD_COND_INITIALIZER)|
> | {|
> | EnterCriticalSection(&ptw32_cond_list_lock);|
> ||
> | cv = *cond;|
> ||
> | /*|
> | * Close the gate; this will synchronize this thread with|
> | * all already signaled waiters to let them retract their|
> | * waiter status - SEE NOTE 1 ABOVE!!!|
> | */|
> | if (sem_wait(&(cv->semBlockLock)) != 0)|
> | {|
> | return errno;|
> | }|
> ||
> | /*|
> | * !TRY! lock mtxUnblockLock; try will detect busy condition|
> | * and will not cause a deadlock with respect to concurrent|
> | * signal/broadcast.|
> | */|
> | if ((result = pthread_mutex_trylock(&(cv->mtxUnblockLock)))
> != 0)|
> | {|
> | (void) sem_post(&(cv->semBlockLock));|
> | return result;|
> | }|
> ||
> | /*|
> | * Check whether cv is still busy (still has waiters)|
> | */|
> | if (cv->nWaitersBlocked > cv->nWaitersGone)|
> | {|
> | if (sem_post(&(cv->semBlockLock)) != 0)|
> | {|
> | result = errno;|
> | }|
> | result1 = pthread_mutex_unlock(&(cv->mtxUnblockLock));|
> | result2 = EBUSY;|
> | }|
> | else|
> | {|
> | /*|
> | * Now it is safe to destroy|
> | */|
> | *cond = NULL;|
> ||
> | if (sem_destroy(&(cv->semBlockLock)) != 0)|
> | {|
> | result = errno;|
> | }|
> | if (sem_destroy(&(cv->semBlockQueue)) != 0)|
> | {|
> | result1 = errno;|
> | }|
> | if ((result2 =
> pthread_mutex_unlock(&(cv->mtxUnblockLock))) == 0)|
> | {|
> | result2 = pthread_mutex_destroy(&(cv->mtxUnblockLock));|
> | }|
> ||
> | /* Unlink the CV from the list */|
> ||
> | if (ptw32_cond_list_head == cv)|
> | {|
> | ptw32_cond_list_head = cv->next;|
> | }|
> | else|
> | {|
> | cv->prev->next = cv->next;|
> | }|
> ||
> | if (ptw32_cond_list_tail == cv)|
> | {|
> | ptw32_cond_list_tail = cv->prev;|
> | }|
> | else|
> | {|
> | cv->next->prev = cv->prev;|
> | }|
> ||
> | (void) free(cv);|
> | }|
> ||
> | LeaveCriticalSection(&ptw32_cond_list_lock);|
> | }|
> | else|
> | {|
> | /*|
> | * See notes in ptw32_cond_check_need_init() above also.|
> | */|
> | EnterCriticalSection(&ptw32_cond_test_init_lock);|
> ||
> | /*|
> | * Check again.|
> | */|
> | if (*cond == PTHREAD_COND_INITIALIZER)|
> | {|
> | /*|
> | * This is all we need to do to destroy a statically|
> | * initialised cond that has not yet been used
> (initialised).|
> | * If we get to here, another thread waiting to initialise|
> | * this cond will get an EINVAL. That's OK.|
> | */|
> | *cond = NULL;|
> | }|
> | else|
> | {|
> | /*|
> | * The cv has been initialised while we were waiting|
> | * so assume it's in use.|
> | */|
> | result = EBUSY;|
> | }|
> ||
> | LeaveCriticalSection(&ptw32_cond_test_init_lock);|
> | }|
> ||
> | return ((result != 0) ? result : ((result1 != 0) ? result1 :
> result2));|
> |}|
>
next prev parent reply other threads:[~2002-09-20 2:40 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2002-09-19 9:17 Michael Johnson
2002-09-19 19:40 ` Ross Johnson [this message]
2002-09-19 9:31 Michael Johnson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3D8A8A88.3000109@ise.canberra.edu.au \
--to=rpj@ise.canberra.edu.au \
--cc=michaelj@maine.rr.com \
--cc=pthreads-win32@sources.redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).